Sensors data fusion for smart decisions making: A novel bi-functional system for the evaluation of sensors contribution in classification problems

Abstract

Sensor fusion has gained a lot of attention during the recent years. It is used as an application tool in different fields including semiconductor-, automotive-, medicine industries. However, finding the right sensor combination for the dedicated application is still very challenging. In this paper, we focus on applying the sensor fusion concept in reference to the prototype-based learning for object classification purposes. In fact, we present a bi-functional system architecture. The system has the feature to evaluate each sensor’s contribution in a predefined classification task. The developed system will preserve the effort and the time spent by engineers to collect a huge quantity of preprocessed samples from each sensor and to try different training configurations. Our approach consists of training a model. The model learns both the predefined classes and additional parameters that represent the contribution of each sensor used in the fusion system for fulfilling the predefined classification task. We illustrate the functionality of our developed system by referring to two different application scenarios. Results validate the dual functionality of our approach as well as the simplicity of the integration of our evaluation system in any further fusion application regardless sensors inputs and classification outputs.